Sewer rehabilitation planning under uncertainty

The need for renewal of sewer systems will increase in coming decades. However, efficient and effective rehabilitation involves substantial planning challenges. These challenges are due to the typically limited degree of documentation of existing infrastructure, variability of input data and deterioration processes, lack of knowledge on the current and indeed future structural condition of sewer systems and unpredictable external influences. Planning challenges are even more pronounced in small utilities as (i) the rehabilitation demand will increase more markedly and (ii) informational, personal and financial planning resources are typically more restricted. Furthermore, it is necessary to make the planning process transparent as a wide range of stakeholders is involved and affected by the planning decisions. The aim of the present study was to develop and identify methodologies for sewer deterioration modelling and hydraulic design that improve (strategic) rehabilitation planning. At the same time the importance of factors affecting the outcome of rehabilitation, including a range of uncertainties, were identified. Estimating the aging behavior of sewer systems is difficult due to the frequently encountered discrepancy between common model structures and data. Existing sewer condition records serving as calibration data are often affected by changes, or even discarded in connection with rehabilitation measures realized in the past. This explains, for instance, the fact that condition records of pipes that have been replaced in the past are no longer available. Consequently, the remaining data no longer exclusively represent the deterioration of pipes but also reflect the (condition improving) rehabilitation of the network. Rehabilitation exerts a selective effect on the pipe population as fast aging pipes are, as a rule, rehabilitated at a lower age. Naively calibrating a deterioration model with such data consequently results in overestimated sewer lifespans. We therefore combined a sewer deterioration model with a rehabilitation model corresponding to the processes reflected by the data. We used Bayesian inference to fully exploit the available information and to overcome problems with parameter identifiability. The combined deterioration and rehabilitation model effectively compensates for the distorting effect of rehabilitation reflected by the data. The identifiability of the model is, however, limited and is only possible when considering prior knowledge. Bayesian inference consequently constitutes a very attractive methodology in the domain of sewer deterioration modelling as it copes with the limited informational content of the data and because prior knowledge is available. Beside the structural aging of sewers, rehabilitation is determined by hydraulic aspects. Current hydraulic design practices have been questioned due to anthropogenic impacts on extreme precipitation properties and, consequently, on the hydraulic performance of sewer systems. We applied stochastic downscaling to refine the precipitation output of multiple climate models to address this impact, including its prediction uncertainty. The data and methodologies applied represent our best knowledge on current and future (extreme) precipitation properties. This information was then used in long-term hydraulic simulations to quantify the performance of selected sewer systems under current and future climatic conditions. The stochastic precipitation data was also used to specifically consider the uncertainties encountered in designing sewer systems. The methodology developed for deterioration modelling and hydraulic design and their combined application was demonstrated in a case study. At the same time we explored different factors influencing sewer rehabilitation as well as the outcomes of various rehabilitation strategies in terms of costs, conditions and hydraulic performance under different socio-economic conditions. Multi-criteria decision analysis (MCDA) was applied to identify well-accepted rehabilitation strategies for given preferences of stakeholders. The impact of climate change and inherent uncertainties were found to be insignificant in the Swiss case studies investigated and the 40-year horizon considered. Significant uncertainty in the extreme precipitation properties arises because we can only observe a ‘random’ realization of the precipitation. This is due of the relatively ‘short’ observation periods of typically 30-40 years. Rehabilitation needs primarily arise from structural deficits and to a lesser extent from hydraulic deficits in the - typical for Switzerland - hydraulically robust sewer networks investigated. However, to attain a higher hydraulic reliability under consideration of these uncertainties relevant extra costs may arise. This is particularly the case if the hydraulic rehabilitation is planned to be completed within a few decades as many pipes would have to be replaced which are still in good structural condition. The results constitute a valuable basis for the revision of current design practice, ensuring a greater hydraulic robustness of sewer systems. The results show that the overall conditions, particularly of the small, “young” aged networks, are distinctly dynamic, non-intuitive and strongly influenced by external factors. This underlines the usefulness of sewer deterioration models and consideration of different scenarios in sewer rehabilitation planning as well as the need for regularly re-evaluating rehabilitation strategies. The most extensive rehabilitation strategies turned out to be the most accepted ones as a result of the MCDA. The decision support obtained by MCDA is, however, supposedly affected by significant biases in the preferences elicited. This is probably due to the failure to provide adequate attributes comprehensible for stakeholders on the basis of the technical and partly probabilistic predictions.